Anon

72 The Basics of financial economeTrics

The multiple regression model is^21

=α+β

β

β

β+e

Mortgage spread (Averageswap rate)

+(10-year/2-yearswapspread)

+(10-year/2-yearswapspread)

+(Swaption volatility)

1

2

3 2

4

Two years of data were used to estimate the regression model. While the
R^2 for the estimated model is not reported, Figure 3.2 shows the actual
mortgage spread versus the spread projected by the regression model for the
Fannie Mae 30-year mortgage passthrough security, one type of mortgage-
backed security (MBS).
Let’s see how the model is used. The analysis was performed
in early March 2004 to assess the relative value of the MBS market.

(^21) See “Mortgages—Hold Your Nose and Buy,” UBS Mortgage Strategist, 9 (March
2004): 15–26. UBS has argued in other issues of its publication that with this par-
ticular regression model the richness of mortgages may be overstated because the
model does not recognize the reshaping of the mortgage market. Alternative regres-
sion models that do take this into account are analyzed by UBS but the results are
not reported here.
Mar-
2002
80
90
100
110
120
130
140
150
160
Actual
Model
Basis Points
May-
2002
Jul-
2002
Sep-
2002
Nov-
2002
May
2003
Jul-
2003
Sep-
2003
Nov-
2003
Jan-
2003
Mar-
2003
Jan-
2004
Mar-
2004
FiGuRE 3.2 Mortgage Spreads: Actual versus Model
Source: Figure 4 in “Mortgages—Hold Your Nose and Buy,” UBS Mortgage Strate-
gist, 9 (March 2004): 19. Reprinted with permission.